[
https://issues.apache.org/jira/browse/HIVE-28532?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
Denys Kuzmenko updated HIVE-28532:
----------------------------------
Labels: hive-4.1.0-must pull-request-available (was:
pull-request-available)
> Map Join Reuse cache allows to share hashtables for different join types
> ------------------------------------------------------------------------
>
> Key: HIVE-28532
> URL: https://issues.apache.org/jira/browse/HIVE-28532
> Project: Hive
> Issue Type: Bug
> Security Level: Public(Viewable by anyone)
> Components: Logical Optimizer
> Affects Versions: 4.0.0
> Reporter: Ramesh Kumar Thangarajan
> Assignee: Ramesh Kumar Thangarajan
> Priority: Major
> Labels: hive-4.1.0-must, pull-request-available
>
> Map Join Reuse cache allows to share hashtables for different join types.
> For example lets take Outer join and Inner join. We cannot reuse a hash table
> for a non-outer join vs outer join. Because outer join cannot accept the hash
> table kind other than HASHMAP, whereas there are other types like HASHSET and
> HASH_MULTISET. Below is the exception when we share the hash table for outer
> join and inner. May be in certain cases we might produce wrong results as we
> expect the hash table to be one type whereas we get the hashtable of another
> type.
> {code:java}
> Caused by: java.lang.ClassCastException: class
> org.apache.hadoop.hive.ql.exec.vector.mapjoin.fast.VectorMapJoinFastStringHashMultiSetContainer
> cannot be cast to class
> org.apache.hadoop.hive.ql.exec.vector.mapjoin.hashtable.VectorMapJoinHashMap{code}
> For this the query plan should be of the form below:
> {code:java}
> Map 11 <- Map 10 (BROADCAST_EDGE)
> Map 5 <- Map 10 (BROADCAST_EDGE)
> Map 9 <- Map 10 (BROADCAST_EDGE) {code}
> where Map 10 gets broadcasted to different mappers and at the same time the
> join type in the Map 11/Map5/Map9 were different.
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
